Ranking of Sources of Uncertainty in the SSMRP Seismic Methodology Chain: Seismic Safety Margins Research Program

摘要

The Seismic Safety Margins Research Program has produced a systems-oriented probabilistic computational procedure for evaluating the seismic safety of nuclear power plants. The Zion, Illinois, nuclear power plant, chosen as typical, served as the subject of Phase I of the program. To estimate the responses of power plant subsystems to earthquake exciation, best-estimate models of the soil, structures, and three piping systems were developed, and key parameters were considered as random variables to represent uncertainties. The variability in seismic input motion was represented by using 90 sets of three-component synthetic time histories. Uncertainties in the computed dynamic responses were estimated by sampling from the input probability distribution functions using a Latin hypercube sampling technique. To separate the influences of the uncertainties in the different links of the seismic methodology chain (seismic input, soil-structure interaction, computation of structure response, and computation of subsystem response) on the final uncertainty, a series of evaluations was performed, with different input parameters varied in each run. Subsystem uncertainties had the strongest impact on the uncertainties in subsystem response. In some cases, the structural uncertainties had a major impact, but is was usually less than that of the subsystem uncertainties. Seismic input uncertainties (given a nominal peak ground acceleration) had a modest impact on subsystem response uncertainties. Soil property variations had little effect at this site; however, these variations are not believed to adequately reflect the total uncertainty in soil-structure interaction calculations for all sites.